ANOVA or t-test?

I want to test 2 different treatments (diets), repeating each treatment 3 times on 25 individuals independent samples. After each repetition, four variables are measured on the individuals (weight, size, variable3, variable4) all quantitative.

I want to assess the differences between the application of diet 2 and diet 1 but I am not totally clear on how to proceed. Statistics is not my field of expertise. I thought I can apply an ANOVA test for each measured variable (weight, size......) and the average of the three repetitions for all the 25 individuals. In this way I will have 4 ANOVAs, each for measured variable between the averages of each treatment.

I have been suggested to use the t-test instead, comparing directly both treatments using the three repetitions, but I am not sure if I should average all the 25 values.

Which test should I use?

Any suggestion is well received.
first of all, to clarify something. A t-test is just a special case of anova (i.e. an anova between two factor levels rather than more than 2). I suppose you can run a dependent sample t-test but you can probably do a little more to refine your analysis.

That said, what this looks like to mean is a repeated measures study. In which case, you can probably do a repeated measure MANOVA design. Do you have any of the participant characteristics included as covariates? If that is the case, a MANCOVA might be more appropriate (for example, Body fat percentage or gender seem like reasonable covariates).
To me this seems to be a mixed design;both between subject(diets) and within subject(the variable).In my view I think mixed-between and within subject analyis or simply spanova can be used